This article appeared in 2000 in the Journal of the Southwest 42:3, 611-33; it is reprinted here with permission.
Stephen A. Marlett
All natural human languages ever described have revealed something about the amazing faculty known to us as "language". All humans are born with this ability and they develop it (unless tragically hindered) with amazing speed. Plus all languages have more complexity to them than scientists have been able to account for with any theory or combination of theories. The Seri language is no exception, clearly demonstrating that there is no such thing as a primitive language, and no relationship between material culture development and language complexity.
Linguists have been pointing out the tragedy that has been taking place in this century as more and more languages have become extinct, one unfortunate side effect of globalization and national education programs. As a result, many languages are no longer available as tools of communication or contemplation) or as objects of research to understand the human mind. Fortunately, the dire (if sometimes wishful) predictions about the demise of the Seri people have been just as inaccurate as those made about the Seri language. And so it has been the privilege of a few of us to study this language during the past fifty years and discover some of its intricacies, its complexities, and its beauty.
We may ask, what is there that is special about Seri? What has been learned during these studies? Of course, there is no reason that a language has to be unique in order to be important, since language serves (among other things) to convey the history, knowledge, culture, and worldview of a people. The Seri language has done this for the many centuries that it has been spoken by a relatively small group of people in the Sonoran desert. Seri is important, first of all, because people speak it. But Seri, like other languages, does indeed have some very special characteristics. Some of these are like treasures of natural history they are just there, and if we open our eyes to see them, we can appreciate them. But others are of a type that are only revealed during the process of relating these facts to linguistic theories. These facts may not be quite so obviously interesting as the first type, but they are usually of even greater importance to linguists. In this paper I present some of these interesting items (both of the natural history type and the theory-relevant type) in a way that, I hope, will be understandable to non-linguists. Seri is not unique in all of these areas, but these are aspects of the language that have made us work hard to describe it accurately and well, and/or they are topics that should make linguists sit up and take notice.
I should point out that the Seri language, while certainly belonging to some ancient language stock of the Americas (commonly claimed to be Hokan), had not been written or been formally studied during all of the centuries that it has been spoken, until the coming of Edward and Mary Moser. This young couple, with only basic training in descriptive linguistic theory and linguistic fieldwork, provided the first phonetic description, phonemic analysis, phonotactic analysis, alphabet based on an analysis of the language, morphological analysis, syntactic analysis, extensive vocabulary, texts, book of stories, and historical accounts recorded in the Seri language. In the space of thirty years, Seri had changed from being a lesser-known language of North America one wordlist was collected around 1850 because a popular belief in Mexico was that it was a dialect of Arabic to being one about which a very great deal is known.
Seri is of extra importance in that it is the sole surviving member of its immediate linguistic family. In fact, we dont even know the names of or anything about other members of this family, nor how many there were. Other languages that might be related to Seri by being part of the putative Hokan super-family (the Yuman languages of Baja California, California, and Arizona; Karok, of Oregon; the Pomoan languages of California; and Chontal of Oaxaca; and others) are distantly related, if at all. For example, whereas Russian and English and Spanish (among many others) are related (coming from a common parent language, proto-Indo-European), evidence of their relationship is readily available and not disputed by anyone. This is not the case for Seri where the evidence of relationship to other languages is not abundant or clear, pace claims otherwise that appear in the literature.
The following sections present eleven areas of the Seri language that investigators have found complicated, tantalizing, interesting, and important.
Seri has one of the most highly developed vocabulary for kin terms that has been described to date. In fact, I believe it may have the most such terms (more than fifty primary terms); the difficulty in determining this for the record book is the availability of information from other languages.
The system is more complicated than many others in that it often distinguishes between the sex of both ego and the sex of the referent, and even someone in between. For example, there are four words for grandparent (unlike three in English, all compound terms: grandparent, grandfather and grandmother): amaz mother of father, act mother of mother, apaz father of father, aaz father of mother. The words for parent include: ata mother (of woman or man), am father (of woman), ai father (of man).
The system is also complicated by the fact that it distinguishes between older and younger referents. For example, there are several words for sibling (unlike only two common ones in English: sister and brother): anyáac his older brother, amáac her older brother, azcz his younger brother, acaz her younger brother, axíiha his or her older brother, apáac his older sister, azáac her older sister, acóome his younger sister, atcz her younger sister, and oyácj his or her sibling.
In addition to these complexities, a term systematically has more than one meaning in many cases. For example, the term azáac means her older sister and also her great-granddaughter, the daughter of her parents older sibling, and her parents great-grandniece.
The Seri language has a rich lexicon, which is remarkable for not incorporating many loanwords. The latter is due primarily to the long isolation from Spanish-speaking communities and some mutual animosity between Seri speakers and Spanish speakers over the centuries. Instead of borrowing, the primary strategy has been to use native words in newly coined fixed expressions. (This shouldnt surprise anyone, really, since English does this all the time also: shock absorber, ice maker, windshield wiper.) Many expressions for things derive transparently from verbs and are in common, everyday use. Examples include ziix icáacötim blanket (lit., thing that one covers up with); ziix haa tiij coos radio (lit., thing that sitting there sings), and ziix cola hapáh flag (lit., thing that is put up high). The language has no difficulty in accommodating new-to-the-culture items, including sports and automobile parts, richly illustrating the creative nature of human language. 
The language also includes a large number of postposition-plus-verb combinations that are reminiscent of the verb-plus-preposition combinations in English, such as put up with to mean endure. The verb cahíti means begin to do, but iti cahíti, with the word iti on it preceding the verb, means fasten, connect. These combinations have to be included in the dictionary since they often have unpredictable meanings. The dictionary also includes a large number of other idiomatic expressions, such as iisax com quiho, lit., to see ones spirit, which means to be agile; and iij cöquiisax, lit., to have spirit elsewhere with respect to (someone), which means to worry about (someone).
As in some other languages of the Americas, the verbs include facets of meaning in them that makes it difficult to translate them with simple glosses. In Seri, these are meanings embedded in the roots themselves, not in some prefix that is added to a basic root. For example, verbs from the domain of carrying include caazi and cahéectim transport, carry; quiip transport on the head, quiiztim transport on hip, quicséenpx and casóompx transport long thing under the arm, cooi transport in one trip, csanj transport person on ones back, cahásquim transport in boat or car, quixop transport with pole over the shoulder, cacáix transport with pole, quizni transport using a handle, and quizcápxla transport under arm.
Seri also has complex morphology. The language can combine various prefixes and suffixes to a root in order to make new words. The verb word has ten places for prefixes, and a few suffixes may also occur. The combination processes are relatively productive, and so any given verb root may occur in hundreds of different words. Because of the regularity, most of these words do not appear in the dictionary as separate items. Therefore the number of actual main entries in the dictionary will number less than 6,000, which appears small when compared to Spanish or English. But the dictionaries of these languages list many derived words as main entries, while this would not be necessary or appropriate in Seri. An English dictionary includes run, runner, and running; the Seri dictionary includes only cpanzx, from which the equivalents of all of the above can be regularly formed, just like runs can be formed from run in English.
Linguists usually describe the words of a language in terms of their syllables. A word such as backstroke in English, for example, consists of two syllables which are phonetically Consonant-Vowel-Consonant (b-a-k) and CCCVC (s-t-r-o-k). All languages have at least syllables that are CV (like to in English), and Seri is no exception. The word cosi thorn has two CV syllables in it (phonetically, k-o and s-i).
The syllable structure of English allows for three consonants at the beginning (as in stroke) and three consonants at the end (as in glimpse), with the option of an extra one at the end of a word (as in glimpsed, which phonetically ends m-p-s-t).
While English has a somewhat more complicated syllable structure than many languages, even the syllables that it has are rather tame in that they generally move from less sonorant consonants (like t) to more sonorant consonants (like r) before arriving at the vowel; true (t-r-u) is a possible word but rtu is not.
Seri allows for more complex syllables, including ones in which there are two or three vowels in the same syllable.
|czxoc||who hacks it off|
|cxtamt||(what is) abundant|
|caai||who makes it|
|caii||who wakes up
|caaol||(what is) pleated|
|caanpx||who returns home|
|quitxl||what breaks into small pieces|
|captj||(what is) wide|
Words may consist of one meaningful part, or more than one. A noun like table has one unit of meaning one "morpheme" and tables has two (the root and the plural suffix). Sometimes a morpheme has more than one form, and these different forms are called "allomorphs". For example, the plural morpheme in English has three common allomorphs which are similar to each other and which may be described as being related to each other in an obvious way: the s in tables which is z-like in pronunciation, the s in cups which is s-like in pronunciation, and the es in bushes.
Besides such allomorphs that are very much like each other in terms of the sounds that they are composed of, one also may find "suppletive allomorphs." Suppletive allomorphs are not phonetic or phonological variations of a basic form; one has to memorize them. One might even be tempted to treat them as entirely different affixes, and perhaps one would not be wrong to do so. Examples of suppletive allomorphs of plural in English include im which occurs on some words of Hebrew origin such as cherubim, i on a few Latinate words such as foci, and en on a few words (oxen; brethren, children).
Seri has very interesting cases of suppletive allomorphy and it has a lot of them. Like other languages, some of these are determined by the phonological context (but not influenced by that context). For example, if one is going to make an imperative in Seri, one has to know whether the verb begins with a short low vowel (a or e), because the prefix is then known to be c/qu, or a back or long low vowel (o, oo, aa, ee) and the verb is intransitive (see "Transitivity"), because the morpheme is then going to be a vowel change rather than a prefix, and otherwise it will be h (with an i preceding the h for other reasons). Some examples:
|oocta||hoocta||Look at it!|
|sanj||ihsánj||Carry him/her on your back!|
Other cases of suppletive allomorphy are determined by morphological factors the presence or absence of other morphemes. For example, even for the imperative morpheme mentioned previously, if the first person singular direct object prefix (hpo me) is present on the verb, the "usual" rules dont apply, and one simply omits any overt realization of the imperative prefix. For example, rather than c for the imperative for the verb tattoo, and rather than h for the imperative of look at and carry on back, the forms simply have null.
|oocta||ihpóocta||Look at me!|
|sanj||ihposánj||Carry me on your back!|
In other cases, the suppletive allomorphy is determined by syntactic factors, the most notable one being that of transitivity (discussed later). Transitivity is partly relevant for the imperative morpheme, but it is very relevant for other prefixes. For example, there is a special prefix that a speaker can use to emphasize his or her opinion, in lieu of the usual first person singular subject agreement prefix on the verb. This "emphatic first person singular" prefix, which occurs in a different position in the word than the "normal" first person singular prefix, has two shapes: aa when the clause is transitive (has a direct object) and caa when it does not. (These prefixes sometimes show up superficially with short vowels or without their vowels.)
|a. Normal first person
Ihsmáhitaha. I wont eat it.
|b. Emphatic first person
Hatée smaahitaha. As for me, I wont eat it.
|c. Normal first person
Ihspíiaha. I will taste it.
|d. Emphatic first person
Hatée sapíiaha. As for me, I will taste it.
|a. Normal first person
Ihpsmóohitimaha. I wont be eating.
|b. Emphatic first person 
Hatée somcóohitimaha. As for me, I wont be eating.
|c. Normal first person
Ihpsfítaha I will stand up.
|d. Emphatic first person
Hatée scafítaha. As for me, I will stand up.
Anyone who tries to learn the Seri language will quickly find out that one complexity is the formation of plurals, whether of nouns or of verbs (since verbs show the number of the subject). Unlike languages that either do not have any indicator of number (some Zapotec languages in Oaxaca, for example), or just add the suffix m to the noun (as some neighboring Uto-Aztecan languages in Sonora, such as Yaqui), or usually add s (like English), Seri flourishes here. In fact, every noun and verb has to have its plural listed in the dictionary because one simply has to learn it.
In Seri, one could ask for a couple of dozen words at random and never see a common way for plurals to be formed. Some examples are given below.
Verbs are similar in that they have different forms depending on whether the action was done by one person or more than one, and whether the action was repeated or not repeated (roughly speaking). And these forms display about the same kind of complexity as the nouns. A few examples are listed here.
|7.||Singular Subject||Plural Subject|
One of the areas in which languages differ is whether the verb agrees with a nominal in the clause. In Spanish, for example, the endings on the verb differ depending on the subject: hablo (yo), hablas (tú), habla (ella, él, usted), etc., as English does to a lesser extent: (I) am, (you) are, (she/he is); (I, you) eat, (she/he) eats. Languages also differ with respect to how many nominals they agree with: none (e.g., Chinese, Zapotecan languages, Mixtecan languages), one (e.g. most Indo-European languages), two (e.g., Nahuatl, Mixe-Zoquean, and Mayan languages, like a majority of languages of the world, apparently), three (more rare), or even four (some Bantu languages; an extremely rare pattern). Seri illustrates the case where agreement is with three nominals. The agreement prefixes are shown in 8.
|8.||Subject||Direct Object||Indirect Object|
|1st person sing.||h, hp||him||he|
|1st person plural||ha||hizi|
|2nd person sing.||m||ma||me|
|2nd person plural||ma||mazi|
|3rd person sing.||(null)||co|
|3rd person plural|
One of the prominent topics in American linguistics in the 1970s was the way in which certain semantic roles such as Addressee and Recipient were encoded in the grammatical structure of a language. In English, for example, one can say Mary gave the book to the librarian, or Mary gave the librarian the book. In both sentences, the librarian receives the book, and hence is the Recipient. But the grammatical structure of the two sentences is different, although English is still rather simple in that the choice is between a simple noun phrase (the librarian) and a prepositional phrase (to the librarian), and the order in which phrases are combined (you cant say Mary gave the book the librarian). One question in linguistics has been how should these facts be accounted for. English isnt the most complicated language in this regard, but it is more complicated than Spanish, for example, where the Recipient is always an indirect object (a prepositional phrase and/or a certain kind of pronoun).
Seri is more complicated in this regard than any language that I have seen so far. Depending on the verb, the Recipient or Addressee may be a direct object (and determine direct object agreement on the verb, as in 9a), or an indirect object (and thus determine indirect object agreement on the verb, as in 9b), or something analogous to a prepositional phrase in English (a kind of postpositional phrase, as in 9c).
a. Zixcám quih him miye.
b. He haamx.
c. Mino hyoomx.
This three-way choice is sometimes complicated by a condition; some verbs only allow singular indirect objects, and so the verb and sentence are quite different syntactically depending on whether one is saying to him/her or to them. This interesting and complicated set of facts (barely described here there is much more to the story) continues to wait for adequate description within a theory of syntax.
Passive clauses are well known to speakers of Indo-European languages such as English and Spanish. They are a grammatical construction in which a verb typically a transitive one in English may occur with a noun phrase in subject position which would be an object in an active clause. So in a passive clause such as the wall was painted by three men, the word wall is subject, but in the active clause three men painted the wall, the word wall is the direct object.
The passive construction of English has been called a personal passive because the subject is one of the participants in the active counterpart. Some languages are also said to have impersonal passives, because they differ from the more familiar passive clauses in just this way. In Indo-European languages such as German, Dutch, and Lithuanian, an intransitive verb may also occur in a passive clause; the result is something like (given literally, to show how it might be in English) it was skated on the ice by the children.
I have already pointed out that Seri has passive clauses. It also has impersonal passive clauses, but they are different from those of German, Dutch and Lithuanian, however, in that they can never be based on intransitive verbs. They are based on transitive verbs and occur when the "pivot" nominal in the clause is plural. To illustrate, one says in Seri the equivalent of I was bitten, you (sg.) were bitten, he/she was bitten with personal passive clauses, rather like in English. But for the plurals one must use the impersonal passive, which would come out in English like it/there was bitten us, it/there was bitten you, and it/there was bitten them.
I dont know whether Seri is the only language in the world that has impersonal passive clauses based on transitive verbs, but it is the only one that has been described as such in the literature.
We usually think of something like transitive and intransitive as terms which some English teacher invented to make life difficult for eighth-grade students. While they may be important in English grammar, many of us would be hard pressed to explain how. Nevertheless, these abstract terms have great significance in Seri; in fact, sometimes I have felt that the language is obsessed with them.
To understand this, let us consider a simple verb root in Seri, ap, which means sew a basket. This verb root may occur in a clause in which a specific basket is mentioned as the direct object, such as she sewed this basket. Or it may occur in a clause in which no specific basket is mentioned, such as she is basket-sewing today. The verb in the first case, with the direct object, is transitive; the verb in the second case, with no direct object, is intransitive. This is extremely important to know in Seri. For example, if one uses the subject I with these verbs, the prefix on the transitive verb is h, but hp on the intransitive. If these verbs are put into the infinitive form, to use in sentences such as I want to sew this basket and I want to basket-sew, the difference between transitive and intransitive is again important; the prefix on the infinitive is iha if the verb is transitive, and ica if it is intransitive. There are several other situations like these where one simply has to know the difference between transitive and intransitive. This knowledge is part of what every speaker of Seri knows subconsciously but very decidedly.
Passive clauses add another interesting twist to this topic. What kind of verb occurs in a passive sentence such as this basket was sewn last year which has a specific basket not as direct object but as subject? While in English grammar, again, it makes no difference, in Seri one cannot remain neutral. And the answer is unambiguous by all of the tests known: passive clauses are intransitive.
This answer is similar to the one reached by examining other (unrelated) languages that have relevant facts, and linguists are not surprised by it. But there is still one more area where the facts of Seri provide some clear answers. Consider a verb such as give which expects two kinds of objects, usually called direct and indirect objects. And suppose that we put such a verb into a passive construction such as I was given the fish. Now we have a passive verb that appears to have a direct object, and the only linguist that (to my knowledge) that discussed this construction in these terms referred to such a clause as a transitive passive clause. And from inspecting English, who could argue otherwise? However, in Seri, the facts are again unambiguous: these clauses are intransitive.
In 1967 a new term appeared in the vocabulary of linguists: switch reference. Using this term, William Jacobsen (1967) described something that he had seen in slightly different ways in a number of languages of North America but which had never been identified as something of theoretical interest. Now it has been found all around the world. Seri is one of the languages that has a switch reference system. The basic idea is that a change in subject from clause to clause is signaled by a little word (often at the end of the clause, as in Seri), as shown in 10. In this sentence, the change from the subject I to the subject we is redundantly signaled by the ta at the end of the first clause, and the switch back to I before the last clause is signaled again by ta. This marker does not occur at the end of the second clause because the third clause has the same referent for its subject as the second clause does.
Hehe pnaacoj pac ihpáaclojta
hacáiin hant_hapáhtoj zo hsáaiaha.
If a sentence is in the past tense, the indicator for switch reference in Seri is ma.
Pedro quih taahitma
ical quih ziix zo cöiyáaha.
In these respects, Seri is rather similar to many other languages with switch reference systems. It is when we look at the system more carefully that an important difference is seen. In Seri, the subject that is relevant is not the surface subject (the one that is marked on the verb), but rather something a bit more abstract. This can be only partially explained here, because of the complexity. An important clue to the nature of the Seri system is found when passive clauses are examined. Passive clauses in Seri, as in many languages, do not permit the overt expression of the Agent. So while a sentence parallel to the bread was eaten is possible, a sentence like the bread was eaten by the bird is not. However, it is precisely this "unexpressed Agent" which is relevant for the switch reference system. The following example illustrates this. The two clauses have the same grammatical subject on the surface; the person who was defeated is the person who left. But the two clauses have different Agents; the person who defeated is different from the person who left. And this is why switch reference marking occurs.
Interestingly enough, the most detailed formal account of switch reference systems that has appeared in the theoretical literature explicitly ruled out this kind of system. The Seri facts remain as a challenge to the dominant current linguistic theories.
Seri has eighteen consonants and eight vowels (four long and four short). Verb roots may begin with almost any of these phonemes. Verbs conjugate very differently when the root begins with a vowel than when it begins with a consonant. To illustrate (and oversimplify), consider the verbs for run and return home in various forms (which I wont bother to translate).
|xompanzx||xomaanpx||Negative of Emphatic Realis|
Despite such clear patterns, there are about a dozen verbs that do something very different. The verb be shiny, to lighning is one such verb:
|14.||be shiny, to lightning|
|xommamjö||Negative of Emphatic Realis|
What is going on here? Note that if the prefix ends in a consonant, the root begins with an identical consonant. If the prefix ends in a vowel, the root begins without such a consonant. And in cases of suppletive prefixes, the one that is appropriate for a consonant-initial verb is chosen. Two similar analyses have actually been proposed, but the basic idea is that these roots begin with a special kind of consonant which is actually none of the regular consonants of the language (such as p, t, c, etc.). Instead, it is a consonant that is lacking any inherent phonological features; it is phonetically empty. But when it follows a consonant, it takes on the "color" of that consonant just as a chameleon takes on the color of its surroundings. But when it follows a vowel, it cannot pick up any features from its surroundings, so it cannot be pronounced; instead, there is simply a hiatus and the speaker moves from the vowel of the prefix to the vowel of the root without any consonant between them.
These facts have played a role in theoretical discussions relating to the question of the abstractness of phonological representations.
In Seri there are several words corresponding to the in English: the most common ones are quih, ac, coi, cop, quij, com, tintica, timoca, tanticat, and tamocat. Some of these words are used only when the word is plural (coi, tanticat, and tamocat). Therefore one can say haxz cop for the dog, but one would say haxaca coi for the dogs.
The choice between the other words for the usually has to do with the position that the item is found in, either temporarily or typically. For example, the word haxz dog can be used with almost any of the words for the, but they are appropriate only as follows:
|15.||haxz cop||(the dog is standing)|
|haxz quij||(the dog is sitting)|
|haxz com||(the dog is lying down)|
|haxz tintica||(the dog is leaving)|
|haxz timoca||(the dog is approaching)|
|haxz quih||(the position of the dog is unknown or irrelevant)|
In some cases, the word for the is almost pre-selected because of the shape. A living tree usually takes the word cop because trees are standing. A crawling insect typically takes the word quij because they are low to the ground and not obviously standing (they dont topple over). A fish typically takes the word com because it is long and dont appear to be standing or sitting. 
The situation becomes more complex when less visible and abstract nouns are considered. And at this point we see the beginnings of a noun class system somewhat analogous to the masculine-feminine noun class system of Spanish, for example, where one has to simply memorize the fact that the word mesa table is feminine and the fact that the word mapa map is masculine. The fact that the word zaah when it means sun takes quij (seated) isnt hard to understand (it is like a ball), but the fact that the same word in its meaning day takes cop (standing) is not understood by examination or reflection. In fact, abstract nouns, including derived words such as iipca rain (from the verb root apca to rain) and iháapl cold weather (from the verb root aapl to be cold) typically take the "standing" word for the, as do liquids (like hax water) and gases (like hai air). Other words showing the beginnings of a system that is not directly tied to observable physical orientation include the following: iisax his/her pirit (takes com), piest fiesta (takes quij), and cöicósim ones smile (takes quij).
We have taken a quick tour around the Seri language in the areas of phonology, morphology, syntax, and lexicon. Other areas (phonetics, semantics, discourse structure, poetics, etc.) will require another day. We have seen that Seri displays a richness and intricacy of structure that will continue to tease and challenge linguists. We hope that it will not be long before a Seri speaker herself or himself will be able and interested to continue these preliminary studies, or start new ones. Much remains to be learned.
 A well-written popular introduction to this topic is Pinker (1995). [Return]
 These points are made repeatedly in modern linguistics textbooks. For a brief discussion, see Lyons (1970, pp. 17ff). [Return]
 See Hale, et al (1992). Languages have become extinct in the past as well because of the extension of political empires and other reasons. [Return]
 The statement by Corey in the 1922 National Geographic was harsh, but certainly didn't represent a minority view: "One need not mourn the Seri Indians too much. It may be that another year will see the last of them" (p. 469). [Return]
 See Langdon (1974). [Return]
 It is possible that Adamo Gilg, Jesuit missionary to the Seri area in the late seventeenth century, actually learned and analyzed the language as he claimed (Di Peso and Matson 1965); no direct evidence is available. [Return]
 For full references, see the bibliography of the Summer Institute of Linguistics available on the internet under "Seri": http://www.sil.org/acpub/biblio/. Interest in Seri in the century preceding 1950 was primarily limited to its place within the kaleidoscope of languages of North America for which relationships in families and super-families were being proposed. With the advent of anthropological and descriptive linguistics that focused on the languages themselves, this began to change. Schweitzer did a preliminary phonemic analysis of Seri, from tape recordings made by Edward Spicer, for a 1953 University of Arizona master's thesis; the lack of direct fieldwork made it difficult to obtain a reliable analysis. [Return]
 McGee (1898:95). [Return]
 Besides technical publications on the phonology and grammar, more popularly accessible works include a small bilingual "vocabulary" (Moser and Moser 1961) that will be superseded by a comprehensive bilingual dictionary (see Moser and Marlett 1997, 1998a-c, 1999a-b for examples of the work in progress), a contribution to the archive of languages of Mexico (Moser with Marlett 1996), a contribution to an intercontinental dictionary series (Moser and Marlett, forthcoming), texts (Moser and Marlett 1998d, for example), songs (Astorga de Estrella et al. 1998) and various ethnobotanical publications culminating in a comprehensive ethnobotany (Felger and Moser 1985). Other items are being published on the Seri page of the SIL Mexico website: http://www.sil.org/americas/mexico/seri/. [Return]
 See Marlett and Moser (1989) and Moser and Marlett (1993). [Return]
 See the information on expressions for Seri names auto parts at http://www.sil.org/americas/mexico/seri/ 12i-Seri-AutoParts.htm. [Return]
 See Moser and Moser (1965) and Marlett (1988). [Return]
 Some theories of syllable structure disallow the possibility of more than one or two vowels in a single syllable. The Seri facts have not yet been accommodated in such theories. [Return]
 These facts are discussed in Marlett (1981a). The following letters have pronunciation different from Spanish: x is a fricative like Spanish and Seri j, but farther back in the mouth; z is a fricative similar to English sh; l is a lateral fricative similar to Welsh ll. [Return]
 The prefix caa-, like other vowel-final prefixes, loses its vowel when it precedes another vowel. [Return]
 See especially Moser (1961). [Return]
 See Siewierska (1998:12ff). The statistics that Siewierska uses apparently do not distinguish between "agreement", "incorporated pronouns", and phonologically dependent pronouns. [Return]
 A quirk of the system does not allow the direct and indirect object prefixes to co-occur. Instead, when one expects them to co-occur, a interesting hybrid form occurs. This is discussed in Marlett (1981a, 1990). [Return]
 See Marlett (1990, 1993). The direct and indirect object agreement markers are written as separate words in the practical orthography (making them resemble Spanish clitic pronouns), but they are technically best analyzed as prefixes. [Return]
 Spanish has a similar construction to that of English, as well as a reflexive passive that I will not discuss here. [Return]
 The impersonal passive also occurs when an instrumental noun phrase (and various others) occur in the clause. One says it/there was hit me with a stick rather than I was hit with a stick. [Return]
 See Marlett (1984a) and Postal (1986). [Return]
 See Hockett (1958:204ff). [Return]
 These facts are treated in Marlett (1981a) and come up in various papers (Marlett 1984a, 1984b, for example) that deal with specific analyses. [Return]
 The basic facts were described in Moser (1978b). Further facts were developed in Marlett (1981a, 1984a, 1984b). [Return]
 See Finer (1985a, b); and see Farrell, Marlett and Perlmutter (1991) for a reply. [Return]
 Rounded consonants have a limited distribution for reasons that are irrelevant here. No root begins with a rounded consonant in Seri. A couple of other consonants are also uncommon. [Return]
 Again, I oversimplify slightly. [Return]
 See Marlett (1981b) and Stemberger and Marlett (1983). [Return]
 See Moser (1978a) and Moser and Marlett (1994a). [Return]
 These words for the are all known to have evolved recently from various "positional" verbs. For example, The verb meaning to be seated has the root -iij. A common form of this verb is quiij, which is the origin of the word quij the (seated). [Return]
Astorga de Estrella, María Luisa; Stephen A. Marlett; Mary B. Moser; and Fernando Nava L. 1998. Las canciones seris: una visión general. Cuarto Encuentro Internacional de Lingüística en el Noroeste, Tomo 1, volume 2, ed. Zarina Estrada Fernández, Max Figueroa Esteva, Gerardo López Cruz, Andrés Acosta, 499-526. Hermosillo: Editorial Unison.
Corey, Herbert. 1922. Adventuring down the west coast of Mexico. National Geographic 42(5): 459-504.
Di Peso, Charles C. and Daniel S. Matson. 1965. The Seri Indians in 1692 as described by Adamo Gilg, S. J. Arizona and the West 7:33-56.
Farrell, Patrick, Stephen A. Marlett, and David M. Perlmutter. 1991. Notions of subjecthood and switch reference: Evidence from Seri. Linguistic Inquiry 22:431-56.
Felger, Richard S. and Mary B. Moser. 1985. People of the desert and sea: Ethnobotany of the Seri Indians. Tucson: University of Arizona Press.
Finer, Daniel. 1985a. The syntax of switch reference. Linguistic Inquiry 16:35-55.
. 1985b. The formal grammar of switch reference. Garland, New York. Ph.D. diss., University of Massachusetts, Amherst, 1984.
Hale, Ken; Michael Krauss, Lucille Watahomigie and Akira Yamamoto, Colette Craig, Laverne Masayesva Jeanne, and Nora England. 1992. Endangered languages. Language 68:1-42.
Hockett, Charles F. 1958. A course in modern linguistics. New York: Macmillan.
Jacobsen, William. 1967. Switch-reference in Hokan-Coahuiltecan. Studies in southwestern ethnolinguistics, eds. D. Hymes and W. Bittle, 238-63. The Hague: Mouton.
Kroeber, A[lfred] L. 1931. The Seri. Southwest museum papers, No. 6. Highland Park, Los Angeles.
Langdon, Margaret. 1974. Comparative Hokan-Coahuiltecan studies: A survey and appraisal. The Hague: Mouton.
Lyons, John. 1970. Noam Chomsky. New York: Viking Press.
Marlett, Stephen A. 1981a. The structure of Seri. Ph.D. diss., University of California at San Diego.
.1981b. The abstract consonant in Seri. Proceedings of the Berkeley Linguistics Society, 154-65.
. 1984a. Personal and impersonal passives in Seri. Studies in Relational Grammar 2, eds. David M. Perlmutter and Carol Rosen, 217-239 . Chicago: University of Chicago Press.
. 1984b. Switch-reference and subject raising in Seri. Syntax and Semantics 16: The syntax of Native American languages, eds. E.-D. Cook y D. Gerdts, 247-68. New York: Academic Press.
. 1988. The syllable structure of Seri. International Journal of American Linguistics 54: 245-78.
. 1990. Person and number inflection in Seri. International Journal of American Linguistics 56:503-41.
. 1993. Goals and indirect objects in Seri. Work Papers of the Summer Institute of Linguistics, University of North Dakota Session 37:1-20.
Marlett, Stephen A. and Mary B. Moser. 1989. Terminología de parentesco seri. Anales de Antropología (Mexico City: UNAM) 26:367-88.
Marlett, Stephen A. and Joseph P. Stemberger. 1983. Empty consonants in Seri. Linguistic Inquiry 14:617-39.
McGee, W. J. 1898. The Seri Indians. Seventeenth annual report of the Bureau of American Ethnology to the Secretary of the Smithsonian Institution. Washington, D.C.: Smithsonian Institution.
Moser, Edward W. 1961. Number in Seri verbs. M.A. thesis, University of Pennsylvania, Philadelphia.
Moser, Edward W. and Mary B. Moser. 1961. Vocabulario seri. Mexico City: Instituto Lingüístico de Verano.
. 1965. Consonant-vowel balance in Seri (Hokan) syllables. Linguistics 16:50-67.
Moser, Mary B. 1978a. Articles in Seri. In Proceedings of the 1977 Hokan-Yuman languages workshop, University of Utah, June 21-23, 1977. Occasional papers on linguistics 2:67-89. Carbondale: Southern Illinois University.
. 1978b. Switch reference in Seri. International Journal of American Linguistics 44:113-120.
Moser, Mary B. and Stephen A. Marlett. 1993. Seri kinship terms. Work Papers of the Summer Institute of Linguistics, University of North Dakota Session 37:21-36.
. 1994a. El desarrollo de clases nominales en seri. Estudios de lingüística y sociolingüística, eds. Gerardo López Cruz and José Luis Moctezuma Zamarrón. Hermosillo: Universidad de Sonora and Instituto Nacional de Antropología e Historia.
. 1994b. Los números en seri. II Encuentro de lingüística en el noroeste, memorias, ed. Zarina Estrada Fernández, II:63-79. Departamento de Letras y Lingüística, División de Humanidades y Bellas Artes, Universidad de Sonora, Hermosillo. [Available on-line at: http://www.sil.org/silewp/1997/005.]
. 1996. Seri de Sonora. Archivo de lenguas indígenas de México. Mexico City: El Colegio de México.
. 1997. Seri dictionary: people and kinship terms. Work Papers of the Summer Institute of Linguistics, University of North Dakota Session, vol. 41. [http://www.und.nodak.edu/dept/linguistics/wp/1997Moser-Marlett.pdf].
. 1998a. Seri dictionary: body parts, bodily processes, sickness and medicine. Work Papers of the Summer Institute of Linguistics, University of North Dakota Session, vol. 42. [http://www.und.nodak.edu/dept/linguistics/wp/1998Moser-Marlett-Body.pdf].
. 1998b. Seri dictionary: earth, sea, sky, time and weather. Work Papers of the Summer Institute of Linguistics, University of North Dakota Session, vol. 42. [http://www.und.nodak.edu/dept/linguistics/wp/1998Moser-Marlett-Earth.pdf].
. 1998c. Seri dictionary: mammals. Work Papers of the Summer Institute of Linguistics, University of North Dakota Session, vol. 42. [http://www.und.nodak.edu/dept/linguistics/wp/1998Moser-Marlett-Mammals.pdf].
. 1998d. How Rabbit Fooled Puma: A Seri Text. Studies in American Indian Languages: Description and Theory, pp. 117-129, edited by Leanne Hinton and Pamela Munro. University of California Publications in Linguistics 131. Berkeley-Los Angeles, University of California Press.
. 1999a. Seri dictionary: colors. Work Papers of the Summer Institute of Linguistics, University of North Dakota Session, vol. 43. [http://www.und.nodak.edu/dept/linguistics/wp/1999Moser-Marlett-Colors.pdf].
. 1999b. Seri dictionary: Plants. Work Papers of the Summer Institute of Linguistics, University of North Dakota Session, vol. 43. [http://www.und.nodak.edu/dept/linguistics/wp/1999Moser-Marlett-Plants.pdf].
. Forthcoming. Seri. Intercontinental Dictionary Series, gen. ed. Mary Ritchie Key.
Pinker, Steven. 1995. The language instinct: The new science of language and mind. London: Penguin.
Postal, Paul M. 1986. Studies of passive clauses. Albany: State University of New York Press.
Schweitzer, Marjorie M. 1953. A provisional analysis of Seri, a native language of Sonora, Mexico. Masters thesis, University of Arizona, Tucson.
Siewierska, Anna. 1998. On nominal and verbal person marking. Linguistic Typology 2(1): 1-55.